Thursday, 23 March 2017

Dimensional Modeling in the Retail Industry: Target Corporation


Target Corporation is one of America’s largest retailers offering everyday merchandise and essentials at a discounted rate to consumers. The firm has around 1,800 outlets across the country with three types of stores, namely, Target - the discount store, SuperTarget - the hypermarket and their flexible format stores – CityTarget and TargetExpress. Target, formerly known as Goodfellow Dry Goods / Dayton’s Dry Goods, was founded by George Dayton in 1902 in Minneapolis, Minnesota. It initially formed a division of the Dayton-Hudson corporation which was the parent company for many departmental store chains. By 1970s, Target was the highest-earning division of the corporation and in 2000, Dayton-Hudson Corporation was renamed as Target Corporation. The firm’s largest competitor is Walmart, which is the largest discount store retailer in the United States and follows the “always low prices” business strategy. Target focuses on attracting the younger demographic by offering a customer centric shopping experience.

Target Corp. focuses on growth by factoring in the latest changes in the retail industry and sticks to its policy of customer prioritization. Performance metrics are one of the most important Key Performance Indicators (KPIs) for any company to evaluate its functional efficiency and growth. Target Corp. is a major player in America’s retail industry. Target’s Chairman and CEO, Brian C. Cornell acknowledged the progress made by the firm over the past few quarters and is keen on being consistent in terms of strategic priorities for growth which have clearly been successful. He also wants to incorporate improvements to enhance this growth further. Identifying and tracking the right performance metrics and implementing strategic policies or changes that align with these metrics would be the ideal approach to achieve this goal. The most important domains for performance metrics evaluation at Target Corporation are:
1.      Inventory Management
2.      Financial Services
3.      Retail Sales
4.      Order Management
5.      Supply Chain Logistics

These domains translate into or help in identifying the Business Process clearly in the 4-step process. The most important performance metrics that would be of interest to the CEO are:
1.    Sales Margin: This is the most important indicator of a retail store’s performance. It is a key indicator of generated sales across a specific period/year. This can be used to evaluate the trends in sales across the country at various locations, across product categories or even Target’s various store types. It can be used in formulation of custom marketing strategies based on location/product category.
2.      Sales Quantity: The number of products sold is a crucial factor in determining a product’s demand in the market and evaluating how well it is being received by customers, especially if it is a new or innovative product. It is a direct indication of profit as well. It is also an essential component of the inventory management system, to maintain the stock level according to a product’s demand at a specific location. It can also be used as a parameter while finalizing decisions in terms of discounted products, stock clearance sales etc.
3.      Average Sales per Transaction (AST):
AST = (total sales/total transactions)
The AST value is an important metric to evaluate the company’s advertising, marketing and promotion campaigns. It is also used in the industry to compare sales at each location and for benchmarking of locations with respect to one another.
4.      Total Sales Dollar Amount: The dollar value of total sales revenue generated in a day, at a store location, is a measure that is useful not only by itself as a performance indicator but is a key component in many KPI calculations. This is often used in conjunction with other metrics to compute various KPIs.
5.      Total COGS Dollar Amount: The COGS or Cost of Goods Sold is one of the key metrics which, when combined with the Total Sales Dollar Amount, provides key insights about a product or store location’s performance and an aggregation of these across product categories or across locations and maybe even a year-wise aggregation provides a strong baseline for judging the financial performance of the company.

Target Corporation is a Retail Industry giant, generating a large amount of data from supply chain, logistics data, inventory management data and transaction data.
·         The information generated at the POS (Point of Sales) at each location everyday can be captured through dimensional modelling techniques and this can be utilized to improve business processes based on key insights and metrics drawn from the data.
·         Dimensional modelling would enable storing large amount of transaction data in a Data warehouse and divide this data into meaningful data marts to perform specific calculations or generate real-time reports to monitor daily sales.
·         Dimensional modelling supports efficient query processing. The de-normalized approach that is followed in dimensional modelling ensures that information access is quick, with high processing speeds in terms of data querying.
·         Extensibility is another important feature of dimensional modelling that would easily accommodate the inclusion of attributes. Any such changes would be seamlessly incorporated in the reports or dashboards as well.
·         Dimensional modelling allows transaction data to be stored at the lowest possible level of granularity. Target can choose to track every transaction at a POS at each store location on any given date. The date dimension plays a critical role in such models.

A Periodic Dimensional model with Transaction fact table would be ideal for Target. This is because it would provide the data as a summary over time such as daily, weekly, monthly or yearly basis which would be ideal for computations.


A sample Dimensional Model for Target’s Retail Sales, specifically for POS, is illustrated below.



References:

Tuesday, 14 February 2017

Business Intelligence Blog

Business Intelligence and Analysis Products

In today’s data-driven world, the importance of Business Intelligence tools is magnified with their role in driving crucial business decisions. Presented below is a comprehensive comparison and weighted analysis of the 5 business tools listed below.
·         Sisense
·         IBM Cognos BI
·         Looker
·         BIRST
·         TIBCO Spotfire


  1.     Sisense

Sisense is a leader in the realm of Business Intelligence software, with the capability to handle large datasets and draw meaningful insights. It has a single-stack architecture, enabling organizations to integrate and visualize disparate datasets instantly. The single-stack architecture also implies that it provides an end-to-end solution in the form of an agile business intelligence solution, right from data preparation to visualization. No additional tools are required, supporting instant deployment. It is particularly well-suited for non-technical users with its easy to use interface.

Strengths
·        Ability to collate data from various data sources quickly without any price constraints
·         Extremely user-friendly, agile system with a drag and drop interface and quick results.
·        Highly efficient system providing data processing speeds 10 times faster than other systems.
·        All-in-one solution with ETL functionality, data warehousing and visualization capabilities.
·         Ideal for handling Big Data, enabling easy analysis of large, complex and scattered datasets.
·    Reduces costs involved by eliminating the need to purchase additional tools and cutting down on resources required to maintain the solution.

Weaknesses
·     Advanced technical expertise is required for setting up the Elasticube feature, which is the crux of Sisense.
·        Reporting feature which extracts and displays dashboard information as a shareable file (like pdf) is not available.
·         User support is free only for a limited period, after which an upgrade is required.
·         Default settings of certain dashboards need to be set manually for each new dashboard.

2.     IBM Cognos 

The IBM Cognos Business Intelligence Suite comprises of a wide range of products, offering services such as Dashboard development, analytics, reporting and scorecarding features as well. It is a web-based product suite but can be used in the offline mode as well. The various modules that make up the Cognos BI Suite ensure flexibility to the users based on their needs. It is a solution that can be used by corporate giants as well as smaller firms by choosing relevant modules.

Strengths
·         Attribute-driven data modeling suggestions, convenient reporting for non-technical users.
·         Information/ dashboards can be accessed from a wide array of devices through mobile app.
·       Highly scalable system with report scheduling capabilities.
·       Analysis and collaboration on content and reports is possible in a single dashboard.
·       Exhaustive documentation, trainings and easily accessible customer support.

Weaknesses
·        Speed of the system is relatively slower and dashboards can take significant time to load.
·       User Interface is not as advanced or easy to use when compared with other BI products.
·         Integration with other products such as Oracle is complex.
·         Error reporting feature is not well developed.

3.  Looker

Looker is a new-age data discovery and analytics platform, making information easily accessible across the organization. It is an advanced web-based product, ideal for data exploration. It helps companies derive maximum value from the data that is collected by enabling them to utilize this data efficiently in different departments.

Strengths
·         No coding skills are required for reporting or dashboard creation.
·        Custom models can be developed within the platform using its proprietary LookML modeling.
·        Browser-based platform enables fast deployment in just a few hours.
·        Real-time database querying ensures transparency between expected and obtained results.
·        Easy integration with third-party applications.
·        Reports and data can be accessed from any device.

Weaknesses
·         Does not support OLAP functionality.
·         Visualization is not as interactive or appealing when compared to other BI products.
·         Data exploration feature is very slow and row limit of 5000 is constraining.
·        The UI is unintuitive, cannot be customized and is not user-friendly.

4. BIRST

Birst is a SaaS Business Intelligence cloud-based platform that is well-suited for enterprises. With its automated data-load feature and a two-tier architecture supporting easy integration of data from various sources, it has the capability to interlink data from across the organization and leverage this data for analytics. It allows decentralized users to modify data models virtually, ensuring that the data security is maintained.


Strengths
·        User-friendly interface with top-notch data integration ability.
·       Automated data-load and refinement into a single layer enables easy data aggregation.
·       Features self-service analytics and can also be integrated with Tableau, Excel and R.
·       Wide array of customizable visualization options.
·       Provides a complete BI environment right from data source integration to analytics.

Weaknesses
·        Cannot be accessed via mobile or other devices, limited to browsers.
·        Steep learning curve for usage of advanced features.
·        Requires improvements in documentation and product support.
·        New versions are not stable and sometimes produces erroneous results.

5. TIBCO Spotfire

Looker is a new-age data discovery and analytics platform, making information easily accessible across the organization. It is an advanced web-based product, ideal for data exploration. It helps companies derive maximum value from the data that is collected by enabling them to utilize this data efficiently in different departments.

Strengths
·        User-friendly platform for performing extensive research and analytics on large datasets.
·        Allows easy integration and synthesis of big data to draw actionable insights.
·        Works well when integrated with excel for graphics and visualization.
·        Processing time to generate reports is very less even in the case of large datasets.
·   Easy configuration with highly customizable elements, in-memory analytics and location based analytics.

Weaknesses
·        Spotfire coding is not very user-friendly and can be cumbersome while writing expressions.
·        Drill-down feature in charts/graphs is not available.
·        Custom formatting of visuals needed to create robust dashboards is not up to mark.
·        Content and user management is done through the desktop client as opposed to web client.


WEIGHTED CRITERIA ANALYSIS
·         Cost Effectiveness: In an age where start-ups are coming up with innovative software products and solutions with competitive pricing, the value of any software product is diminished if there is a similar product offering the customer same functionality at a lower price. The return on investment for any product is crucial, especially in the case of BI tools which are used to drive decisions that impact the business. Any organization, large or small, would always prefer a cost effective, efficient product with the right features to effectively visualize key insights as opposed to a cheaper alternative with features that don’t tie up with the firm’s objectives. Hence, this criteria accounts for 20%.

·         Data Integration: Data integration is a crucial factor in the case of any BI solution. Statistically significant information with correlations and meaningful insights are usually drawn when data from related processes/ sources is integrated and visualized. Thus, 15% is allocated for this criterion since it is extremely essential for any BI tool to have the capability to bring together data from various disparate sources onto a single platform for further processing or running analytics.

·         Data Modeling and Analytics: The importance of data modeling is realized while trying to extract information according to client requirements and present accurate results drawn from raw data. A BI Tool must provide users with the capability for data slicing and dicing from data sources. The tool must also support drill-down capability to extract information even at a granular level. This further supports accurate analytics based on details from the raw data. A weightage of 30% has been allocated to this criterion since it is the crux of business intelligence.

·         Interactive Visualization: In today’s data driven world, it is of utmost importance to condense all the information and insights gleaned from terabytes of data into a visual format that captures the essence of what would be useful to drive decisions. A BI tool should provide comprehensive data visualization features in order to ensure that the right insights are conveyed to the end user. It is also necessary for reports, scoring charts and graphs to be flexible enough to show how a change in one metric impacts the result. Hence, this criterion is allocated 20% of the weightage.

·         Processing Speed: The technology industry is all about fast results. It is very important for any BI tool to process data and display the output of a query or a report without too much of a lag. A frequent problem that users face while handling large datasets is system crashes. A weightage of 15% is allocated for this factor since a BI tool must be capable enough to handle the sheer volume of data and produce the output without any lag.


Criteria
Weight
Sisense
Cognos
Looker
Birst
Spotfire
Cost Effectiveness
20%
8
9
7
8
7
Data Integration
15%
10
8
8
8
8
Modeling and Analytics
30%
9
8
5
7
9
Interactive Visualization
20%
9
8
6
7
8
Processing Speed
15%
10
7
6
7
8
Points
100%
9.1
8.05
6.20
7.35
8.10
Rank

1
3
5
4
2


Weighted Scoring
After performing the weighted analysis of the 5 BI tools based on the core criteria, Sisense emerges as the clear winner with Looker being on the last spot.

·         Sisense is ranked first primarily due to its advanced data integration capabilities, providing an all-in-one, agile solution with great visualization that provides quick results. The only drawback is the lack of transparency in terms of pricing.

·         TIBCO Spotfire, with its excellent data modeling capabilities, great value for money and eye-catching visualizations, comes in second. The only concern with Spotfire is the cost. It is comparatively costly but does provide value for money in terms of delivering results.

·         Spotfire is closely followed by IBM’s Analytics platform, Cognos. Cognos has been a pioneer in the business intelligence realm. With a variety of products in the IBM Cognos Suite, it provides firms with the option to pick relevant modules. Cognos Express also offers a cheaper alternative with full functionality to smaller firms that may be hesitant to invest in the entire suite. The only drawback when it comes to Cognos is the lag, leading to dashboards loading very slowly on the web interface and complexities involved in integration with non-IBM products.

·         Birst is cost effective as it is an open source tool and easily available to the users. It also has efficient modeling capabilities. But cause of concern is the speed and visualization. 

·         Looker is the lowest ranked BI tool among the 5 chosen BI tools. The principal drawback in Looker is the inability to perform Online Analytical Processing, which is a crucial element in Analytics. Also, data processing speeds are very slow with many bugs and system crashes. Lack of visualization is also a minus, putting it in the last spot among the 5 chosen tools.